PARENTHOOD IN ACADEMIA¶

Europe¶

==================================================================================================================

Involved Researchers¶

Out[1]:

==================================================================================================================

Content¶

  1. Overview
  2. Basic Survey Stats
  3. Scholarly Data
  4. Observational Studies Exploration
  5. Vienna Scientific Cluster
  6. Outlook

==================================================================================================================

I. Overview¶

Research Question:
How does parenthood impact researchers of different academic seniority in Europe?
Reference Study:
Morgan et al. (2021), The unequal impact of parenthood in academia, Science Advances.
Data Source:
Survey data from academics in EU encoding relevant variables received through questions.
Out[2]:

The hypotheses we aim to test are based on survey variables linked to question IDs. An object-oriented pipeline is being developed to make this process reusable. The next section gives an initial look at the collected data.

==================================================================================================================

II. Basic Survey Stats¶

Survey Structure

➤ Part 1: Demographics (self, parents, spouse, children)
➤ Part 2: Parental & Childcare Policies (leave, time use, responsibilities)
➤ Part 3: Academic Experiences (network satisfaction, exit intentions)
The following section presents selected descriptive statistics from the survey. 
===== Survey Timeline =====
Survey ran between 2024-03-25 and 2024-08-20.

===== Response Statistics =====
Total unique researchers (participants): 10072
Note: The Number of responses varies across questions; percentages refer to respondents of a specific question, not the full sample.
Out[5]:
Responded with birth year: 6525 (64.78%)
Did not respond: 3547 (35.22%)
➤ About 50% of respondents were born between 1969 and 1984 (Gen X).
Out[7]:
➤ Slightly more women than men participated in survey; non-binary significantly underrepresented.
Out[8]:
➤ Majority of respondents originate from western European countries; researchers from outside Europe are a minority.
Out[9]:
➤ Over half of the respondents are tenured faculty.
Out[10]:
➤ Over 70% of respondents have partners who are not in academia; only around a quarter have academic partners.

==================================================================================================================

III. Scholarly Data¶

➤ Respondents could provide links to their academic profiles.
➤ These sources are essential to retrieve yearly publications and citations, as proxies for productivity.
Out[11]:
1242 ORCID
1177 CV
928  Google Scholar
401  Research Gate
180  Scopus
103  Web of Science
243  other URLs
7    Open Alex
➤ Started with ORCID and Scopus IDs from respondents.
➤ Used Scopus API to retrieve ORCIDs.
➤ valid ORCIDs were used to retrieve author data from Open Alex API (yearly publications and citation counts, and field of researcher).
➤ Expanding dataset using Google Scholar profiles via SerpApi.
➤ Different structure, requires alignment with Open Alex data.
➤ Google Scholar provides only researcher-defined keyword interests; may be used for inferring research domains.

==================================================================================================================

IV. Observational Studies Exploration¶

Objective

➤ Estimating the effect of parental leave on productivity.

Hypothesis 1

➤ "Regardless of gender, parents get equally penalized if they take parental leave."

Approach

➤ Constructing a causal graph (DAG) to identify confounders that create backdoor paths.
Out[12]:
Backdoor paths identified through DAG construction: 
Parental Leave ← Gender → Productivity
Parental Leave ← Career Stage → Productivity
Parental Leave ← Country → Productivity
Parental Leave ← Career Stage ← Gender → Productivity
Parental Leave ← Career Stage ← Country → Productivity
➤ Backdoor paths involve variables that affect both parental leave and productivity. If we do not account for them, they introduce false associations and bias our results.
➤ This was addressed by exploring matching and two-way fixed effects methods for blocking these paths and isolating the actual impact of parental leave on productivity.

Matching Method

➤ Creating Groups
➤ Treated (525) & Control (140)

➤ Propensity Score Matching
➤ Nearest neighbor with replacement
➤ Matching covariates: Gender, Academic Role

➤ Fit Logit Regression
➤ Compute propensity scores

➤ Standardized Mean Differences (SMDs)
➤ Pre- and post-matching covariate balance check

➤ Estimate ATT
➤ Average Treatment Effect on the Treated

➤ Bootstrapping
➤ Robustness check

Out[13]:

Preliminary Results

➤ ATT (Effect of Taking Leave on Publications):
➤ On average, taking parental leave is associated with a reduction of 14.49 publications
➤ Bootstrapped 95% CI: [–24.05, –4.31]

➤ Gender disparity:

Out[14]:
➤ Penalty observed only for women
➤ Women who took leave published ~22 fewer papers, while men gained ~5

➤ Role-based heterogeneity:

Out[15]:
➤ Tenure-track faculty and those in non-tenure-track or other roles face stronger negative effects
➤ In contrast, postdocs and tenured faculty show a gain in publications after taking leave

Limitations

➤ Matching-based estimates are limited by:
➤ Country and Discipline covariates not controlled for
➤ No temporal dimension; pre/post-leave effects not isolated
➤ Next step: Move to two-way fixed effects models once panel data is complete
➤ Enables longitudinal comparisons stratified by gender, career stage, and discipline
➤ Accounts for both time-invariant individual differences (e.g. discipline, baseline productivity) and common time trends (e.g. pandemic effects)

==================================================================================================================

V. Vienna Scientific Cluster¶

➤ Access to VSC enables faster, large-scale computation.
➤ Application process involved:
➤ Writing project description with justification and ethical conduct.
➤ Providing 2–3 researchers to review and endorse the proposal.
➤ Application currently under review.

==================================================================================================================

VI. Outlook¶

Remaining Work

➤ Populate the two‑way fixed‑effects model with researchers’ complete publication histories
➤ Controls for time‑invariant individual traits and period‑specific shocks
➤ Continue ingesting external sources (ResearchGate, Web of Science, respondent‑supplied URLs)
➤ Parse uploaded CVs
➤ 803 PDF
➤ 261 DOCX
➤ 92 DOC
➤ 21 other (PNG, RTF, Pages, HTML)
➤ Launch follow‑up survey round
➤ Refactor code pipeline and enrich visualisations (e.g. show qualitative answers split by gender)